How to set CMAKE_INSTALL_PREFIX option correctly
It would be advatnegaous for https://github.com/usatlas/analysisbase-dask/issues/4#issuecomment-1806459779 (as pointed out by @alheld) to be able to ship a built version of the extensions with the Docker container image when doing testing so that the module is available as soon as
. /release_setup.sh
is run and no building is done by the end user (critical for distributed runs). We would normally run something in the image build like
RUN . /release_setup.sh \
&& git clone --recurse-submodules --branch py_el_tool_test https://gitlab.cern.ch/gstark/pycolumnarprototype.git \
&& cd pycolumnarprototype \
&& cmake -DCMAKE_INSTALL_PREFIX=<something here> -S src -B build \
&& cmake build -LH \
&& cmake --build build --clean-first --parallel 4 \
&& cmake --build build --target install
but a naive attempt at -DCMAKE_INSTALL_PREFIX=/venv
inside of sslhep/analysis-dask-base:latest
fails. There's a warning given of
CMake Warning at /usr/AnalysisBaseExternals/24.2.26/InstallArea/x86_64-centos7-gcc11-opt/cmake/modules/AtlasInternals.cmake:591 (message):
CPack packaging will only work correctly with
CMAKE_INSTALL_PREFIX=/PyColumnarPrototypeDemo/1.0.0/InstallArea/x86_64-centos7-gcc11-opt
Call Stack (most recent call first):
CMakeLists.txt:61 (atlas_cpack_setup)
but that doesn't subtantially clear things up.
@gstark is it clear to you off the top of your head how we should be installing built tools into the environment?
Edited by Matthew Feickert